fix: base64 decoding

Fix the base64 decoding. If there's an 'A' character in the encoded
text, the base64 decode function returns an INVALID_PADDING error. The
reason lies in the way Base64Decoder.init tries to find a suitable
invalid character. Fix this by defining the invalid character as 0xff
(which is already the case for a decoding without padding).

This error has not been caught by the test harness, because no test
contains an 'A' character in the the encoded text yet. Add a new test.

test/unit/stdlib/encoding/base64.c3

@@ -19,6 +19,7 @@ fn void encode()
         { "foob", "Zm9vYg==" },
         { "fooba", "Zm9vYmE=" },
         { "foobar", "Zm9vYmFy" },
+        { "test", "dGVzdA==" },
     };
     foreach (tc : tcases)
     {
@@ -41,6 +42,7 @@ fn void encode_nopadding()
         { "foob", "Zm9vYg" },
         { "fooba", "Zm9vYmE" },
         { "foobar", "Zm9vYmFy" },
+        { "test", "dGVzdA" },
     };
     foreach (tc : tcases)
     {
@@ -63,6 +65,8 @@ fn void decode()
         { "Zm9vYg==", "foob" },
         { "Zm9vYmE=", "fooba" },
         { "Zm9vYmFy", "foobar" },
+        { "Zm9vYmFy", "foobar" },
+        { "dGVzdA==", "test" },
     };
     foreach (tc : tcases)
     {
@@ -85,6 +89,7 @@ fn void decode_nopadding()
         { "Zm9vYg", "foob" },
         { "Zm9vYmE", "fooba" },
         { "Zm9vYmFy", "foobar" },
+        { "dGVzdA", "test" },
     };
     foreach (tc : tcases)
     {
@@ -95,4 +100,4 @@ fn void decode_nopadding()
         usz nn = b.decode(tc.in, buf[:n])!;
         assert(buf[:nn] == tc.out);
     }
-}
+}